SSC 5073

PERGAMON Solid State Communications 113 (2000) 553–558

www.elsevier.com/locate/ssc

In situ Raman scattering studies of the amorphous and crystalline

Si nanoparticles
A.A. Sirenko a,*, J.R. Fox a, I.A. Akimov a,b, X.X. Xi a, S. Ruvimov c, Z. Liliental-Weber c
a
Department of Physics, The Pennsylvania State University, University Park, PS 16802, USA
b
A.F. Ioffe Physico-Technical Institute, Russian Academy of Sciences, St Petersburg 194021, Russian Federation
c
Lawrence Berkeley Laboratory, Berkeley, CA 94720, USA
Received 10 September 1999; received in revised form 26 October 1999; accepted 24 November 1999 by J. Kuhl

Abstract
We report on in situ studies of the vibrational properties of Si nanoparticles and ultrathin layers grown by dc magnetron
sputtering in ultrahigh vacuum on amorphous MgO and Ag buffer layers. The average thickness of the Si layers ranged from
monolayer coverage up to 200 Å. Transmission electron microscopy has been used to determine size and shape of the Si
nanoparticles. Changes in the phonon spectra of Si nanoparticles during the crystallization process have been studied by
interference enhanced Raman scattering technique. Marked size-dependences in the phonon density of states and the relaxation
of the k-vector conservation with decrease in size of the Si nanoparticles have been detected. The transition between crystalline-
and amorphous-like behavior takes place in the particles with an average number of Si atoms equal to …7 ^ 2† × 102 : q 2000
Elsevier Science Ltd. All rights reserved.
Keywords: A. Nanostructures; D. Phonons; E. Inelastic light scattering

1. Introduction due to a difference in the bulk dispersions and the dielectric

As the size of condensed matter systems is reduced in one
or more dimensions to nanoscale levels, the electron and
phonon states are influenced due to confinement. The
discovery of visible luminescence at room temperature
from porous Si [1,2] increased interest in the modification
of electronic and vibrational properties of silicon nano-
crystals. The prospect of realizing light-emitting devices
stimulated intensive investigations of various nanometer-
sized Si-based structures using different kinds of preparation
methods [2–8]. Recent experimental studies of nanocrystal-
line Si (nc-Si) demonstrated substantial changes in the
luminescence properties and electronic bands, which were
attributed to quantum confinement caused by the restricted
size of the particles and a breakdown of k-vector conserva-
tion [9–12].
Phonon states are also modified in semiconductor nanos-
tructures (or quantum dots) from those of the bulk. Theory
predicts confinement of the optical and acoustic vibrations
[13–15], which can be described macroscopically as being
* Corresponding author.
0038-1098/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved.
PII: S0038-109 8(99)00539-6
properties of the nanoparticles and the surrounding media.
This effect has been experimentally observed in a number of
semiconductor nanostructures by means of Raman scatter-
ing [5,16,17,18,19,20]. However, previous Raman scatter-
ing studies were mostly focused on samples with Si particles
within a matrix or with ligands, such as polycrystalline
hydrogenated Si [5], Si-rich SiO2 films [6], prepared by
pulverizing bulk Si [7], porous Si [17–19,21], or Si prepared
by thermal evaporation [22]. In these systems, the influence
of a matrix or surface contamination on the phonon spectra
cannot be neglected. Thus, in the analysis of the observed
shifts and broadening of the optical phonon lines, it was
difficult to distinguish between effects due to phonon
confinement, modification of the surface states, and
matrix-induced stress. While all the above mentioned effects
are possible, the phonon states of isolated Si nanoparticles
have been relatively unexplored experimentally and publi-
cations on in situ measurements of nc-Si are rare [23].
In this paper, we report on in situ Raman studies of ultra-
small Si particles and thin films grown in ultrahigh vacuum
(UHV). The phonon states of the Si nanoparticles free of
chemisorbed species as well as changes induced by the
554 A.A. Sirenko et al. / Solid State Communications 113 (2000) 553–558
Fig. 1. TEM cross-section image of the recrystallized 30 Å sample
grown on Ag buffer.
crystallization process and hydrogen termination will be
discussed and vibration properties of the Si nanoparticles
grown on dielectric and metallic substrates will be
compared. Results of in situ electron energy loss spectro-
scopy (EELS) will be used for the characterization of the
optical gap of Si nanoparticles.
2. Experiment
Ultrathin layers of Si were deposited on amorphous MgO
or Ag buffers at room temperature by dc magnetron sputter-
ing in a UHV chamber. A power of 6 W/cm 2 was applied to
the Si sputter target. The background pressure of Ar was
kept around 4 mTorr during the Si growth process. Before
deposition of Si, the substrates were annealed at 5008C. The
growth rate of Si, calibrated by both ex situ ellipsometry and
profilometry, was about 0.5 Å/s. The average thickness of
the Si layers, d, ranged from monolayer coverage up to
200 Å. In the following, the Si samples will be identified
by their value of d. To crystallize Si ultrathin layers, the
samples were heated up to about 5008C for 30 min in a
UHV. The temperature of the sample was estimated by
means of both a thermocouple attached to the sample holder
and a pyrometer.
The grown ultrathin Si layers consist of nanometer-size
clusters of a hemispherical shape formed on the substrate.
Their average size and concentration varied with d. Ex situ
transmission electron microscopy (TEM) was employed to
characterize the shape and size of Si clusters in the samples
grown on Ag buffer layers. Ag films with Si clusters were
mechanically removed from the Si substrate, first, and then
folded so that the Si clusters on the edge can be studied by
TEM. Two Si clusters are seen in a cross-section image of
the 30 Å sample annealed at 5008C (Fig. 1). The Si clusters
in this sample have a 70–100 Å base width and are 30–40 Å
high. The Si layer with a thickness of about 10, covers Ag
buffer between quantum dots. These data are in a quantita-
tive agreement with the calibrated average amount of depos-
ited Si. Our results with in situ core-level X-ray
photoemission (XPS) show that at d # 50 A;  Si forms
nanoscale islands on MgO substrates as well. For thicker
layers of Si, the characteristic MgO signal in the XPS spec-
tra quickly disappears, corresponding to a complete cover-
age of the buffer-layer surface with Si atoms. We
determined that the Si clusters contain about …70 ^ 20† × d
(Å) atoms for d less than 50 Å. For the samples with d
more than 50 Å, no nanoparticle formation has been
observed and their properties are the same as of thin
films. It is similar to the growth regime of Ge nanoparticles
with a hemispherical shape investigated previously with the
same apparatus [24].
The Raman measurements were performed in situ at room
temperature and at a pressure of 2 × 10210 Torr in the UHV
system. The main experimental problem in the system of
two-dimensional composition of Si nanoparticles was a
weak Raman signal due to extremely small scattering
volume. For Si nanoparticles grown on MgO layers, we
improved the situation by using interference enhanced
Raman scattering (IERS) [25,26]. The Raman signal from
the Si layer was increased by a factor of 10 in comparison
with similar samples grown on Ag due to constructive inter-
ference in a bilayer structure. We found that in combination
with multichannel detection [27], IERS can be sensitive
even to a submonolayer coverage. Recently this technique
provided useful information about the vibrations of surface
and near-surface atoms of isolated Ge nanoparticles [24] and
carbon nanocrystalline clusters and films [28]. In samples
grown on MgO buffer-layers, the interference enhancement
of the laser electric field close to the surface (the position of
the Si layer) was achieved by utilizing a bilayer structure,
which consists of a transparent dielectric film (MgO) depos-
ited on a metallic layer (Ag) [25,26]. The entire structure
was grown on a Si substrate. The thickness of the amorphous
Ag, which works in a bilayer structure as a light reflector,
was about 4000 Å. This is thick enough to screen comple-
tely any possible Raman signal from the Si substrate. The
MgO thickness of 500 Å was calculated to give optimal
interference enhancement at the exciting wavelength. The
choice for the dielectric buffer layer (MgO) was determined
by the two following reasons. First, the luminescence and
Raman signal of MgO is weak at the green light excitation
and, second, the analysis of the XPS spectra is easier than,
e.g. in the case of a SiO2 buffer, which also contains Si
atoms.
Raman spectra were measured with a resolution of
0.8 cm 21 using a SPEX Triplemate spectrometer equipped
with a charge-coupled device (CCD) detector cooled with
liquid nitrogen. The 5145 Å line (2.4 eV) of an Ar 1-ion
laser was used for excitation. The pumping power
density was about 1 W/cm 2. Both the intensity and
linear polarization of the Raman signal were analyzed
in the conventional backscattering configuration (laser
beam and scattered light are perpendicular to the sample
plane). In the following, symbols VV and VH denote the
scattering configurations with the polarization of the Raman
signal being parallel and perpendicular to the polarization of
the exciting beam, respectively. EELS spectra were
measured with incident electron energy of 50 eV. An
LK2000 spectrometer was operated with 5 eV pass energy
yielding a resolution of approximately 0.03 eV.
 A.A. Sirenko et al. / Solid State Communications 113 (2000) 553–558 555

Fig. 3. The HWHM d TO and position D TO of the TO Raman peak as

functions of the average thickness of the amorphous Si layers.
Closed and open symbols correspond to samples grown on MgO
and Ag, respectively. Dashed lines guide the eye.

Fig. 2. (a) Normalized Raman spectra of amorphous Si layers

measured with excitation at 2.4 eV. The respective average thick-
nesses are given next to the spectra in Å. The spectra were vertically
shifted for clarity. The horizontal solid lines indicate the zero-signal
levels. (b) Raman spectra of the 30 Å sample measured in VV and
VH polarization.

3. Results and discussion

Raman spectra of amorphous samples with a different

average thickness of the Si layer grown on MgO are
shown in Fig. 2(a). As expected, the spectra of the relatively
thick layers are similar to that of amorphous Si (a-Si) and
consist of the TA, LA, and TO strong bands centered around
150, 310 and 480 cm, respectively, and the weak LO
shoulder at 370 cm 21 [29–31], marked in Fig. 2(b). The
Raman spectra are polarized at the linearly polarized excita-
tion. The VV and VH spectra are shown in Fig. 2(b) for the
30 Å sample. The degree of linear polarization, defined as
…IVV 2 IVH †=…IVV 1 IVH †; is about 0.25, which is close to the
value previously determined for a-Si [32]. IVV and IVH are
the Raman intensities in the corresponding configuration.
As the Si layer thickness decreases, the high frequency
half-width-at-half-maximum (HWHM) of the TO band d TO
defined as shown in Fig. 2(a) changes from 35 to 50 cm 21.
At the same time, the position of the maximum D TO moves
Fig. 4. (a) Normalized Raman spectra of the 20 Å sample measured
towards the low Raman shifts. The dependences of D TO and at different temperatures during crystallization process. (b) Spectra
d TO on d are presented in Fig. 3 for samples grown on both of the k ˆ 0 optical phonon in the same sample before (open
MgO and Ag. The observed broadening of the optical band circles) and after termination of the surface with a hydrogen plasma
in the Raman spectra of Si layers corresponds to the modi- (closed circles). Vertical dashed line shows a position of the c-Si
fication of the phonon density of states due to changes in the (100) phonon [35] at 519:7 ^ 0:8 cm21 :
556 A.A. Sirenko et al. / Solid State Communications 113 (2000) 553–558
angular distribution of the sp-bonds in the near-surface
atomic layers and dangling bonds on the surface [24,32].
The surface effects are expected to be more pronounced in
ultrathin Si layers consisting of amorphous clusters, where a
fraction of the surface atoms to interior atoms increases
significantly. Indeed, the significant variation in the phonon
spectra occurs for d less than 50 Å, the same thickness-range
where the formation of the Si clusters was observed. In the
case of relatively thick layers with d $ 50 A;  any possible
corrugation of the Si surface …Dd ˆ ^10 A†  should not
significantly affect the Raman spectra, which are similar
to that of the a-Si films. Note, that in the samples grown
on MgO buffer layers, formation of the Si–O bonds at the
interface could lead to the similar transformation in the
Raman spectra, especially in the ultrathin layers, where
the fraction of the surface atoms is high. However, the thick-
ness dependences of the optical band for the samples grown
on both MgO and Ag were identical (see Fig. 3). Thus, we
conclude that the vibration properties of the amorphous Si
nanoparticles are determined by the size effects. Our experi-
ments are also important to distinguish between size-depen-
dent modifications of the phonon density of states from that
due to contamination of the Si surface with hydrogen or
oxygen. The latter might be neglected for our samples, but
could be very strong in, e.g. porous Si.
After heating the samples, a remarkable transformation in
the Raman spectra was observed. Fig. 4(a) shows several
spectra for the 20 Å sample taken at different temperatures.
The sharp crystalline peak arises when the temperature
increases to more than T . 2608C: Note the temperature-
induced red shift of this Raman line. At the same time, the
intensity of the amorphous-like broad features decreases at
T . 3208C: This indicates a partial recovering of the k-
vector conservation condition with crystallization of the Si
layers allowing only scattering by optical phonons at the
zone center …k ˆ 0†; where the TO and LO phonons in
bulk crystalline Si (c-Si) are degenerate [33]. The degree
of linear polarization of this peak is about 0.16, which corre-
sponds to the random orientation of the crystallographic
directions in the Si nanoparticles. The crystallization
temperature was found to be around 250–3008C for samples
grown on both MgO and Ag (note that the crystallization
temperature for a-Si films varies in a wide range depending
on the sample preparation condition and substrates, see, e.g.
Ref. [34]).
The position of the k ˆ 0 peak does not change signifi-
cantly in different recrystallized samples. Its shape is asym-
metric and a full-width-at-half-maximum (FWHM) varies
from 6 cm 21 for thick layers to about 8 cm 21 (see Fig.4(b))
for 20 Å sample. For d # 50 A;  the typical red shift of this
line with respect to the position of c-Si (100) phonon (the
room temperature value for the k ˆ 0 phonon in c-Si varies
in literature between 519 and 522 cm 21, see, e.g. Refs. [35])
is about 1:5 ^ 0:5 cm21 : From the first glance, this shift
could be attributed to the phonon confinement. However,
after termination of the sample surface with a hydrogen
plasma, the position of the k ˆ 0 peak shifts back and coin-
cides within our experimental accuracy with the c-Si (100)
phonon. The measurements were taken with the same pump-
ing power and the heating effect was excluded. Thus, we
conclude that the initially observed red shift of about
1.5 cm 21 is determined not by confinement of the optical
phonons but is mostly due to the surface strain. In passivated
recrystallized samples, hydrogen reduces the strain at the
nanoparticle surface by destroying unfavorable tensile-
strained Si–Si bonds [36]. A slight decrease in the FWHM
of the Raman line (see Fig.4(b)) can be also due to this
effect. Further we will briefly discuss effects, which can
influence the position of the phonon lines in the Raman
spectra of nanocrystals of intermediate size. After that we
will proceed to the case of ultrasmall crystalline particles.
The absence of the confinement-induced shift of the opti-
cal phonon in the samples with d between 20 and 50 Å lacks
a straightforward interpretation. On one hand, a simple esti-
mation of the confinement-induced frequency shift of the
optical phonon in a hemispherical particle [15] results in a
value of about 5 cm 21 for the sample with d ˆ 20 A;  which
should be measurable with our experimental accuracy. On
the other hand, there are several factors, which can strongly
affect this simple picture decreasing the expected confine-
ment-induced effect below our experimental limits:
1. The wide size distribution of the nanoparticles, which can
result in a preferable excitation of the quantum dots with
the maximum size increasing the average size of the dots,
which are probed in the Raman experiments.
2. The inhomogeneous broadening of the Raman line,
which is caused by the higher-order confined modes
with the frequencies lower than the primary confined
Raman line [37] and by the splitting between TO and
LO phonons.
3. The strain between the dots and the substrate [5,7]. Note
that the compressive strain shifts the line position in the
opposite direction to that induced by confinement.
4. An increase of the fraction of the surface to interior atoms
in smaller nanoparticles, leading to modification of the
phonon density of states [28,32]. Note that the contribu-
tion of the surface vibrations may affect the Raman spec-
trum as well [38].
The combination of the above-mentioned factors, in prin-
ciple, could be responsible for the fact that in the studied Si
clusters, which contain only a few 1000 atoms, no strong
systematic shift of the k ˆ 0 peak was detected. A compar-
ison of higher resolution Raman measurements with precise
theoretical investigation of the phonon confinement, which
should take into account the exact shape of the nanoparticles
(see Fig. 1), surface effects, and interaction with substrate,
are required to clarify the situation. At this point we can only
mention that our data are more similar to results for phonon
confinement in thin Si films, where the relatively small
Raman shifts were observed [15], and to specific cases of
porous Si [39].
 A.A. Sirenko et al. / Solid State Communications 113 (2000) 553–558
Fig. 5. Raman spectra of the 10 Å sample before (solid line) and
after the annealing (dashed line). Inset shows the ratio of the k ˆ 0
peak intensity Ikˆ0 to that of the amorphous background Ia as a
function of the average thickness of the Si layers.
Nevertheless, in the ultrathin recrystallized Si layers, we
expected a formation of much smaller Si clusters with a
typical size of 10–30 Å, where the effect of the phonon
confinement should become dominant. However, instead
of a larger red shift of the phonon line, the k ˆ 0 peak
disappeared in samples with d # 10 A:  As the average
thickness of the Si layer decreases, the relative intensity of
the k ˆ 0 peak to the intensity of the amorphous-like broad
peak at 480 cm 21 becomes smaller. The ratio of the k ˆ 0
peak intensity to the amorphous-like background intensity is
Fig. 6. EELS spectra of the 5 Å sample before (triangles) and after
the annealing (circles). Dashed lines guide the eye to the onset of the
electronic transitions indicated with arrows. Spectra for polycrystal-
line Si and SiO2 are shown for comparison with open squares and
solid line, respectively.
557
shown in inset to Fig. 5. In the case of the relatively thick
films, with d greater than 45 Å, the complete k-vector
conservation could be achieved after annealing, which
manifests itself by the vanishing of the amorphous-like
features in the Raman spectra. In contrast, for Si layers
with d less than 10 Å, the Raman spectra do not change
with annealing and practically no modifications have been
observed even after heating up to 6008C for 1 h. Two corre-
sponding spectra of the 10 Å sample grown on a Ag buffer
layer and measured at room temperature before and after
annealing are shown in Fig. 5. This result is in agreement
with the earlier prediction [32] and experimental observa-
tions [23,38,40] of the fact, that in ultrasmall nanocrystals
the k ˆ 0 line vanishes and Raman spectra of both covalent
and ionic nanocrystals become amorphous-like when they
are small enough. Two reasons can be adduced to explain
this effect. First, even in the ideal case of isolated nanocrys-
tal, the decrease in size breaks the crystal periodicity allow-
ing observation of the scattering from entire branches of the
acoustic and optical phonons. The situation becomes similar
to a system of weakly interacting molecules in liquids or
gases and the consideration of k-vector conservation
becomes needless. Second, the fraction of disordered
surface atoms increases with respect to the ordered atoms
within a crystalline lattice [28,32]. The observed size-
dependent modification of the Raman spectra is important
for determination of the minimum size of a crystalline
system, which can satisfy k-vector conservation conditions.
According to the aforementioned empirical relation between
d and the average size of the nanoparticles, the transition
between crystalline- and amorphous-like behavior takes
place in the recrystallized samples with an average number
of Si atoms in one hemispherical particle equal to …7 ^ 2† ×
102 :
To support this statement an additional verification of the
crystalline structure of the ultrasmall nanoparticles is
required. Note, that one of the alternative explanations of
the observed effect can be an increase of the temperature of
the amorphous-to-crystalline structural transition. However,
our in situ measurements of electron energy losses demon-
strate significant modification of the electronic gap with
annealing in the same temperature range, confirming that
the structure of the annealed Si nanoparticles is crystalline.
The EELS spectra obtained before and after recrystallization
for a d ˆ 5 A  sample grown on SiO2 are shown in Fig. 6.
The intensity was normalized to the elastic peak amplitude.
The onsets of electronic transitions (shown with arrows in
Fig. 6) in the amorphous sample occurs at much lower
energy …0:45 ^ 0:05 eV† than that seen in the annealed
sample …1:55 ^ 0:05 eV; similar values for the optical gap
can be extracted from the EELS spectra by computing the
absorption coefficient and using analogous techniques as
those employed to amorphous semiconductors, for more
detail see Ref. [41]). The latter value is much higher than
the indirect gap of bulk crystals [42], which provides a
clear evidence of the electron confinement in ultrasmall
558 A.A. Sirenko et al. / Solid State Communications 113 (2000) 553–558
nanocrystals. The observed increase in the electronic gap is
in a good agreement with the recently reported changes in
the band ages of silicon nanocrystals by X-ray absorption
and photoemission [11].
4. Conclusions
The phonon states of Si nanoparticles free of chemisorbed
species, which were grown and measured in ultrahigh
vacuum, have been studied. Remarkable transformation in
the phonon density of states with decrease in size of amor-
phous Si nanoparticles was observed. It has been found that
the surface strain induced by the dangling bonds can result
in a red shift of the optical phonon frequency for relatively
big nanocrystals. We consider the observed relaxation of the
k-vector conservation in ultrasmall nanocrystals as the most
important result of this paper. The average number of Si
atoms per one particle, which corresponds to the transition
between crystalline- and amorphous-like behavior, has been
determined. A confinement-induced increase in the electro-
nic gap of ultrasmall Si particles has been measured by
means of in situ EELS providing a clear evidence of their
crystalline structure.
Acknowledgements
The authors are grateful to V.I. Merkulov for useful
discussions. This work was supported by USDOE Grant
DE-FG02-84ER45095 and NSF Grant DMR-96-23315.
References
[1] A.G. Cullis, L.T. Canham, Nature (London) 353 (1991) 353.
[2] A.G. Cullis, L.T. Canham, P.D.J. Calcott, J. Appl. Phys. 82
(1997) 909 and references therein.
[3] R. Tsu, J. Morais, Bowhill, in: Z. Feng, R. Tsu (Eds.), Porous
Silicon, World Scientific, Singapore, 1994, p. 443.
[4] K. Eberl, K. Brunner, W. Winter, Thin Solid Films 294 (1997)
98.
[5] Z. Igbal, S. Veprek, A.P. Webb, P. Capezzuto, Solid State
Commun. 37 (1981) 993.
[6] E.W. Forsythe, E.A. Whittaker, F.H. Pollak, S.W. Sywe, G.S.
Tompa, B.A. Khan, J. Khurgin, H.W.H. Lee, F. Adar, H.
Schaffer, in: R.W. Collins (Ed.), Microcrystalline and Nano-
crystalline Semiconductors, Materials Research Society
Symposia Proceedings No. 358MRS, Pittsburgh, 1994, p. 187.
[7] X.S. Zhao, Y.R. Ge, J. Schroeder, P.D. Persants, Appl. Phys.
Lett. 65 (1994) 2033.
[8] L. Tsybeskov, MRS Bull. 23 (4) (1998) 33 and references
therein.
[9] W.L. Wilson, P.F. Szajowski, L.E. Brus, Science 262 (1993)
1242.
[10] S. Schlupper, S.L. Friedman, M.A. Marcus, D.L. Adler, Y.-H.
Xie, F.M. Ross, T.D. Harris, W.L. Brown, Y.J. Chabal, L.E.
Brus, P.H. Citrin, Phys. Rev. Lett. 72 (1994) 2648.
[11] T. van Buuren, L.N. Dinh, L.L. Chase, W.J. Siekhaus, L.J.
Terminello, Phys. Rev. Lett. 80 (1998) 3803.
[12] D. Kovalev, H. Heckler, M. Ben-Chorin, G. Polisski, M.
Schwartzkopff, F. Koch, Phys. Rev. Lett. 81 (1998) 2803.
[13] R. Alben, D. Weaire, J.E. Smith Jr., M.H. Brodsky, Phys. Rev.
B 11 (1975) 2271.
[14] T. Takagahara, J. Lumin. 70 (1996) 129.
[15] I.H. Campbell, P.M. Fauchet, Solid State Commun. 58 (1986)
739.
[16] A. Ekimov, J. Lumin. 70 (1996) 1 and references therein.
[17] A. Fillos, S.S. Hefner, R. Tsu, J. Vac. Sci. Technol. B 14 (6)
(1996) 3431.
[18] Y. Kanemutsu, H. Uto, Y. Masumoto, T. Masumoto, T.
Futagi, H. Mimura, Phys. Rev. B 48 (1993) 2827.
[19] S. Guha, P. Steiner, W. Lang, J. Appl. Phys. 79 (1996) 8664.
[20] Z. Igbal, S. Veprek, J. Phys. C: Solid State Phys. 15 (1982)
377.
[21] H. Tanino, A. Kuprin, H. Deai, N. Koshida, Phys. Rev. B 53
(1996) 1937.
[22] Y. Gao, T. López-Rı́os, Solid State Commun. 60 (1986) 55.
[23] S. Hayashi, H. Abe, Jpn. J. Appl. Phys. 23 (1984) L824.
[24] J. Fortner, J.S. Lannin, Surf. Sci. 254 (1991) 251.
[25] G.A.N. Connel, R.J. Nemanich, C.C. Tsai, Appl. Phys. Lett.
36 (1980) 31.
[26] W.S. Bacsa, J.S. Lannin, Appl. Phys. Lett. 61 (1992) 19.
[27] J.C. Tsang, in: M. Cardona, G. Guntherodt (Eds.), Light Scat-
tering in Solids, Topics in Applied Physics, Vol. V, Springer,
Berlin, 1989, p. 233, chap. 6.
[28] V.I. Merkulov, J.S. Lannin, J.M. Cowley, in: R.W. Collins
(Ed.), Advances in Microcrystalline and Nanocrystalline
Semiconductors, Materials Research Society Symposia
Proceedings No. 452MRS, Pittsburgh, 1996, p. 225.
[29] J.E. Smith Jr., M.H. Brodsky, B.L. Crowder, M.I. Nathan, A.
Pinczuk, Phys. Rev. Lett. 26 (1971) 642.
[30] J.S. Lannin, P.J. Carrol, Philos. Mag. 45 (1982) 155.
[31] G. Nilson, G. Nelin, Phys. Rev. B 6 (1972) 3777.
[32] J.S. Lannin, in: J.I. Pankove (Ed.), Semiconductors and Semi-
metals, 21(B), Academic Press, Orlando, 1984, p. 159.
[33] P.Y. Yu, M. Cardona, Fundamentals of Semiconductors,
Springer, Berlin, 1996, p. 103.
[34] D. Bimberg, et al., in: O. Madelung, et al. (Eds.), Numerical
data and functional relationships in science and technology,
Landolt-Bornstein, New Series, Group III, Vol. 17i, Springer,
Berlin, 1985, p. 5.
[35] D. Bimberg, et al., in: O. Madelung, et al. (Eds.), Numerical
Data and Functional Relationships in Science And Technol-
ogy, Landolt-Bornstein, New Series, Group III, Vol. 17a,
Springer, Berlin, 1982, p. 72.
[36] I. Kaiser, N.H. Nickel, W. Fuhs, W. Pilz, Phys. Rev. B 58
(1998) R1718.
[37] C. Trallero-Giner, A. Debernardi, M. Cardona, A.I. Ekimov,
Phys. Rev. B 57 (1998) 4664.
[38] T. Okada, T. Iwaki, K. Yamamoto, H. Kashara, K. Abe, Solid
State Commun. 49 (1984) 809.
[39] F. Kozlowski, P. Steiner, F. Lang, in: Z. Feng, R. Tsu (Eds.),
Porous Silicon, World Scientific, Singapore, 1994, p. 149.
[40] S. Hayashi, Solid State Commun. 56 (1985) 375.
[41] G.P. Lopinski, J.S. Lannin, Appl. Phys. Lett. 69 (1996)
2400.
[42] G.D. Cody, in: J.I. Pankove (Ed.), Semiconductors and Semi-
metals, Vol. 21B, Academic Press, Orlando, 1984, p. 11.